Tiny spectrometer enables cost-effective space-borne sensing

Small size and weight, and the ability to survive launch and operate reliably for extended periods in space are paramount concerns when building instruments for satellites and planetary probes. Because of their compact design, waveguide spectrometers seem ideal for space-based spectroscopy. The spectrometer chips are stackable: a rugged, compact instrument can be designed using multiple waveguide slabs for detection of separate spectral targets.¹ However, a typical input aperture is only a few micrometers across, so optical collection efficiency from incoherent, distributed sources (such as the sun) is poor. Increasing the optical throughput would offer significant advantages for space applications.

In bulk optics, Fourier-transform IR (FTIR) spectrometers offer throughput advantages compared to dispersion devices, but they typically involve mechanical scanning of the optical path in a Michelson interferometer. Spatial heterodyne spectrometry (SHS) is a static Fourier technique, without moving parts, based on spatially distributed interferograms formed by diffraction gratings that replace the Michelson mirrors.² A monolithic SHS instrument recently proved useful as a UV spectral imager and the primary payload of the Space Test Program satellite-1 (STPSat-1).³ A near-IR version has also been developed for spatial heterodyne observations of water (SHOW).⁴

In optical waveguides, the incident radiation is dispersed into spectral components, as for example in waveguide echelle and arrayed-waveguide gratings.⁵ A Fourier-transform SHS waveguide spectrometer can be formed by a Michelson-type arrangement of two arrayed-waveguide gratings.⁶ Another simple, practical approach uses an array of Mach-Zehnder interferometers (MZIs) on a silicon chip with an incremental optical-path difference across the array (see Figure 1).⁷

Figure 1. Schematics of the waveguide spectrometer formed by N arrayed Mach-Zehnder interferometers (MZIs). Light enters multiple waveguides that feed individual MZIs, which have incremental optical-path differences ΔL_i (i=1, 2, …, N). Distribution of optical power at output waveguides P^out (i) results in the discrete spatial interferogram that encodes the input spectrum p_in(λ).

Because of the small size and high refractive-index contrast of silicon waveguides, many interferometers can be densely arrayed on a single chip of a few centimeters in size. Such arrays offer a manifold increase of the optical throughput compared to single-input designs. Because of the different periodic transmission of individual interferometers, the spectrum of the incident light is encoded in the output power pattern (see Figure 2). It can be retrieved by performing a discrete Fourier transform of the output power data, similarly to conventional FTIR spectroscopy.

Figure 2. Simulated interferogram formed by 200 MZIs (spectrometer resolution 0.025nm, spectral range 2.5nm). (inset) The input water absorption at 15km altitude is retrieved using apodized Fourier transformation.

Our prototype spectrometer (see Figure 3) is an array of 50 MZIs on a silicon-on-insulator (SOI) chip. The inputs of individual MZIs are closely spaced to maximize light collection, while the outputs are spaced to match a detector array. The chip is designed for 0.4nm resolution and 10nm spectral range at a wavelength of 1550nm. It was fabricated using commercially available SOI wafers with 2.2μm silicon (Si) and 0.4μm of buried Si oxide. We formed ridge waveguides by reactive ion etching of the Si layer and depositing a silica top cladding.

Figure 3. Waveguide layout and fabricated silicon chip of our spectrometer, formed by 50 MZIs (shown against a Canadian two-dollar coin).

We tested the prototype using a tunable laser source and a spectral-source emulator. The output ports were imaged onto an indium gallium arsenide IR camera (see Figure 4) and processed to obtain discrete spatial fringes, from which we retrieved the input spectrum. We observed the spectral signatures of emulated emission lines in the output fringe patterns.

Figure 4. IR images of the spectrometer outputs (interferograms) for different spectral inputs.

In addition to the large optical throughput, an important advantage of our device is that deviations from the ideal design appear as systematic errors in the interferograms. Once the device has been fabricated and fully characterized, the errors can simply be corrected by software calibration. We are currently working to improve spectrometer performance and on signal processing that incorporates data calibration. In the near future, we aim to develop the SHOW application at a wavelength of approximately 1364nm.

Support for this work by the Canadian Space Agency, COM DEV Ltd., the National Research Council of Canada, and the Ontario Centres of Excellence is gratefully acknowledged.